Wiki
Wiki

Research notes on every problem, and a library of the papers behind them. Built from the open erdos repository.

Updated


Source. Theorem 2, p. 423 (with its derivation running onto p. 424), of P. Erdős and A. Szemerédi, On multiplicative representations of integers, J. Austral. Math. Soc. Ser. A 21 (1976), no. 4, 418--427, doi:10.1017/S144678870001925X, as named on the source card.

Statement

Setting (pp. 418, 421). For sequences of positive integers A={a1<a2<⋯ }A=\{a_1<a_2<\cdots\} and B={b1<b2<⋯ }B=\{b_1<b_2<\cdots\}, A(x)A(x) and B(x)B(x) count their terms up to xx, and g(n)g(n) is the number of solutions of n=aibjn=a_ib_j.

Theorem 2 (p. 423). Suppose A(x)>c1xA(x)>c_1x and B(x)>c2xB(x)>c_2x. Then there is an n<xn<x with g(n)>(log⁡x)αg(n)>(\log x)^\alpha.

The exponent α\alpha is a positive constant that the statement does not specify; the paper adds (p. 424) that α≤1/log⁡2\alpha\le1/\log2 is easy to prove and that it is not clear how far this can be improved. The printed statement has an unmatched parenthesis in "g(n)>g(n)> log x)αx)^\alpha" and bounds nn by xx. The outline on p. 420 states the same result as display (8), max⁡n≤x2g(n)>(log⁡x)c3\max_{n\le x^2}g(n)>(\log x)^{c_3}, which p. 421 calls best possible apart from the value of c3c_3, and the derivation on p. 424 counts products of two integers below xx, which range up to x2x^2; the bound n<xn<x in the printed statement is therefore read here as a misprint for n<x2n<x^2 (an observation of this page, not of the paper).

Read depth. Claims checked: the statement, display (8) and the remarks on α\alpha were read clause by clause on the printed pages; the derivation was read but not checked step by step.

Proof pointer

Pages 423--424. The paper calls the theorem an immediate consequence of Erdős's 1960 theorem (On an asymptotic formula in number theory, Leningrad Univ. 15 (1960), 41--49): the products aibja_ib_j number more than a constant times x2x^2, while the distinct integers of the form klkl with k,l<xk,l<x number fewer than x2/(log⁡x)αx^2/(\log x)^\alpha, so some integer has a number of representations of order at least (log⁡x)α(\log x)^\alpha. The paper's p. 421 remark that (8) is best possible takes the aa's and bb's with at most log⁡log⁡n\log\log n prime factors.

Dependencies

Erdős's 1960 theorem on the number of distinct products klkl with k,l<xk,l<x, quoted without proof.

Bears on

None of the corpus's problem pages directly.